Xiangqi Dev Log: Cannons and the River
After two quiet months, the questions about xiangqi development are about progress and release date. This dev log answers both: how the cannon captures, what the river changes, why the soldiers hold up the schedule, and an honest release window rather than an invented date.
Xiangqi development at YiBoard went quiet for about two months while ratings, anti-cheat and the mobile board took the front seat, and the same two questions kept arriving: how much xiangqi progress is actually real, and when is the xiangqi release date. Here is the honest log. The cannon is finished and passes its tests. The river took three rewrites and still catches us out. The soldiers, which look like the simplest pieces on the board, are what holds the schedule. None of it runs anywhere special: it is the same browser board you already use for gomoku on yiboardgame.com, so you can open the play page in another tab and check the parts that are already live.
What is finished, and what is not
The rules engine is done: every legal move for all seven piece types is generated and validated, including the two that most apps get wrong, the blocked horse leg and the elephant that cannot cross the river. The board renders on a phone without scrolling. The AI plays near the strength of a casual club player and runs the same alpha-beta search as gomoku, described in How the engine works. Not finished: English names for the openings, game review, and the rating conversion. We would rather ship a correct board with fewer features than a wide one that misrules a cannon capture.
The cannon: one piece with no Western equivalent
The cannon moves like a rook. It does not capture like one. To take a piece, the cannon has to jump over exactly one other piece, the screen, anywhere along the line. No screen, no capture. Two screens, also no capture. That single rule is why xiangqi stops feeling like chess once the middle game opens, and why a piece that looks useless in the opening becomes the one you watch in the last twenty moves.
- Moves: any distance along a rank or file, like a rook, as long as the path is clear.
- Captures: only by jumping exactly one piece. The screen can be yours or the opponent.
- Zero screens means it takes nothing on that line, even with a target one point away.
- Two or more screens blocks the capture too. This is the case most hobby implementations miss.
- Value drops in the endgame, when the board empties out and screens get scarce.
The river is a rule, not decoration
The board has 90 intersections and 32 pieces, and a line across the middle called the river. It is not drawn there for looks. Three things change depending on which side of it a piece stands on.
- Elephants cannot cross. Their two point diagonal step is also blocked by a piece on the midpoint, so they are defenders on a short leash.
- Soldiers move one point forward only. After crossing the river they also move sideways one point at a time, and never backward.
- Horses cross freely, but the horse leg block still applies: a piece on the adjacent orthogonal point cancels that step.
- The two generals face each other across the river on files 4 and 6, each locked inside a 3 by 3 palace.
What we build next, in order
- English opening names, starting with Central Cannon and Screen Horse.
- Game review that marks the move where the evaluation turned, reusing the review view from gomoku.
- A separate rating scale. Xiangqi will not start at 1200 on the same ladder; the reasons are in Where the ladder comes from.
- Undo and takeback in teaching games, which is the first thing schools ask for.
Is there a xiangqi release date yet
Not a day, and we are not going to invent one. The honest range is a playable board with correct rules and a bot before the end of this year, with ratings and review after it. Two things could push that back: the rating conversion has to be calibrated against real games, and the opening library needs English names somebody can check. If you want the shape of the gap we are closing, The Xiangqi Platform Gap lays it out. Until the board ships, the live part of the site is the gomoku ladder on the rankings page.
FAQ
Is there a xiangqi release date for YiBoard?
No fixed day. The next milestone is a playable board with correct rules, and ratings and review come after it. We publish progress here instead of naming a date we would then miss.
Will xiangqi use the same rating as gomoku?
No, it gets its own scale. The 1200 starting point is calibrated for gomoku games, and putting both games on one ruler would make both numbers meaningless.
What makes the cannon hard to implement?
The screen rule. A cannon captures only when exactly one piece sits between source and target, so the validator has to count pieces on the line instead of checking that the path is empty. Two screens is illegal, and several apps get that wrong.
Can I play xiangqi on yiboardgame.com today?
Not yet. Gomoku is live with no sign-up and no install. The xiangqi board runs the same browser engine and will appear on the same play page once it is ready.